Merge branch 'svn-crlf' of git://bogomips.org/git-svn into ew/svn-crlf
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * See Documentation/technical/api-directory-listing.txt
6 *
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #define NO_THE_INDEX_COMPATIBILITY_MACROS
11 #include "cache.h"
12 #include "config.h"
13 #include "dir.h"
14 #include "attr.h"
15 #include "refs.h"
16 #include "wildmatch.h"
17 #include "pathspec.h"
18 #include "utf8.h"
19 #include "varint.h"
20 #include "ewah/ewok.h"
21 #include "fsmonitor.h"
22
23 /*
24 * Tells read_directory_recursive how a file or directory should be treated.
25 * Values are ordered by significance, e.g. if a directory contains both
26 * excluded and untracked files, it is listed as untracked because
27 * path_untracked > path_excluded.
28 */
29 enum path_treatment {
30 path_none = 0,
31 path_recurse,
32 path_excluded,
33 path_untracked
34 };
35
36 /*
37 * Support data structure for our opendir/readdir/closedir wrappers
38 */
39 struct cached_dir {
40 DIR *fdir;
41 struct untracked_cache_dir *untracked;
42 int nr_files;
43 int nr_dirs;
44
45 struct dirent *de;
46 const char *file;
47 struct untracked_cache_dir *ucd;
48 };
49
50 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
51 struct index_state *istate, const char *path, int len,
52 struct untracked_cache_dir *untracked,
53 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
54 static int get_dtype(struct dirent *de, struct index_state *istate,
55 const char *path, int len);
56
57 int count_slashes(const char *s)
58 {
59 int cnt = 0;
60 while (*s)
61 if (*s++ == '/')
62 cnt++;
63 return cnt;
64 }
65
66 int fspathcmp(const char *a, const char *b)
67 {
68 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
69 }
70
71 int fspathncmp(const char *a, const char *b, size_t count)
72 {
73 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
74 }
75
76 int git_fnmatch(const struct pathspec_item *item,
77 const char *pattern, const char *string,
78 int prefix)
79 {
80 if (prefix > 0) {
81 if (ps_strncmp(item, pattern, string, prefix))
82 return WM_NOMATCH;
83 pattern += prefix;
84 string += prefix;
85 }
86 if (item->flags & PATHSPEC_ONESTAR) {
87 int pattern_len = strlen(++pattern);
88 int string_len = strlen(string);
89 return string_len < pattern_len ||
90 ps_strcmp(item, pattern,
91 string + string_len - pattern_len);
92 }
93 if (item->magic & PATHSPEC_GLOB)
94 return wildmatch(pattern, string,
95 WM_PATHNAME |
96 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
97 else
98 /* wildmatch has not learned no FNM_PATHNAME mode yet */
99 return wildmatch(pattern, string,
100 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
101 }
102
103 static int fnmatch_icase_mem(const char *pattern, int patternlen,
104 const char *string, int stringlen,
105 int flags)
106 {
107 int match_status;
108 struct strbuf pat_buf = STRBUF_INIT;
109 struct strbuf str_buf = STRBUF_INIT;
110 const char *use_pat = pattern;
111 const char *use_str = string;
112
113 if (pattern[patternlen]) {
114 strbuf_add(&pat_buf, pattern, patternlen);
115 use_pat = pat_buf.buf;
116 }
117 if (string[stringlen]) {
118 strbuf_add(&str_buf, string, stringlen);
119 use_str = str_buf.buf;
120 }
121
122 if (ignore_case)
123 flags |= WM_CASEFOLD;
124 match_status = wildmatch(use_pat, use_str, flags);
125
126 strbuf_release(&pat_buf);
127 strbuf_release(&str_buf);
128
129 return match_status;
130 }
131
132 static size_t common_prefix_len(const struct pathspec *pathspec)
133 {
134 int n;
135 size_t max = 0;
136
137 /*
138 * ":(icase)path" is treated as a pathspec full of
139 * wildcard. In other words, only prefix is considered common
140 * prefix. If the pathspec is abc/foo abc/bar, running in
141 * subdir xyz, the common prefix is still xyz, not xuz/abc as
142 * in non-:(icase).
143 */
144 GUARD_PATHSPEC(pathspec,
145 PATHSPEC_FROMTOP |
146 PATHSPEC_MAXDEPTH |
147 PATHSPEC_LITERAL |
148 PATHSPEC_GLOB |
149 PATHSPEC_ICASE |
150 PATHSPEC_EXCLUDE |
151 PATHSPEC_ATTR);
152
153 for (n = 0; n < pathspec->nr; n++) {
154 size_t i = 0, len = 0, item_len;
155 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
156 continue;
157 if (pathspec->items[n].magic & PATHSPEC_ICASE)
158 item_len = pathspec->items[n].prefix;
159 else
160 item_len = pathspec->items[n].nowildcard_len;
161 while (i < item_len && (n == 0 || i < max)) {
162 char c = pathspec->items[n].match[i];
163 if (c != pathspec->items[0].match[i])
164 break;
165 if (c == '/')
166 len = i + 1;
167 i++;
168 }
169 if (n == 0 || len < max) {
170 max = len;
171 if (!max)
172 break;
173 }
174 }
175 return max;
176 }
177
178 /*
179 * Returns a copy of the longest leading path common among all
180 * pathspecs.
181 */
182 char *common_prefix(const struct pathspec *pathspec)
183 {
184 unsigned long len = common_prefix_len(pathspec);
185
186 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
187 }
188
189 int fill_directory(struct dir_struct *dir,
190 struct index_state *istate,
191 const struct pathspec *pathspec)
192 {
193 const char *prefix;
194 size_t prefix_len;
195
196 /*
197 * Calculate common prefix for the pathspec, and
198 * use that to optimize the directory walk
199 */
200 prefix_len = common_prefix_len(pathspec);
201 prefix = prefix_len ? pathspec->items[0].match : "";
202
203 /* Read the directory and prune it */
204 read_directory(dir, istate, prefix, prefix_len, pathspec);
205
206 return prefix_len;
207 }
208
209 int within_depth(const char *name, int namelen,
210 int depth, int max_depth)
211 {
212 const char *cp = name, *cpe = name + namelen;
213
214 while (cp < cpe) {
215 if (*cp++ != '/')
216 continue;
217 depth++;
218 if (depth > max_depth)
219 return 0;
220 }
221 return 1;
222 }
223
224 #define DO_MATCH_EXCLUDE (1<<0)
225 #define DO_MATCH_DIRECTORY (1<<1)
226 #define DO_MATCH_SUBMODULE (1<<2)
227
228 static int match_attrs(const char *name, int namelen,
229 const struct pathspec_item *item)
230 {
231 int i;
232
233 git_check_attr(name, item->attr_check);
234 for (i = 0; i < item->attr_match_nr; i++) {
235 const char *value;
236 int matched;
237 enum attr_match_mode match_mode;
238
239 value = item->attr_check->items[i].value;
240 match_mode = item->attr_match[i].match_mode;
241
242 if (ATTR_TRUE(value))
243 matched = (match_mode == MATCH_SET);
244 else if (ATTR_FALSE(value))
245 matched = (match_mode == MATCH_UNSET);
246 else if (ATTR_UNSET(value))
247 matched = (match_mode == MATCH_UNSPECIFIED);
248 else
249 matched = (match_mode == MATCH_VALUE &&
250 !strcmp(item->attr_match[i].value, value));
251 if (!matched)
252 return 0;
253 }
254
255 return 1;
256 }
257
258 /*
259 * Does 'match' match the given name?
260 * A match is found if
261 *
262 * (1) the 'match' string is leading directory of 'name', or
263 * (2) the 'match' string is a wildcard and matches 'name', or
264 * (3) the 'match' string is exactly the same as 'name'.
265 *
266 * and the return value tells which case it was.
267 *
268 * It returns 0 when there is no match.
269 */
270 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
271 const char *name, int namelen, unsigned flags)
272 {
273 /* name/namelen has prefix cut off by caller */
274 const char *match = item->match + prefix;
275 int matchlen = item->len - prefix;
276
277 /*
278 * The normal call pattern is:
279 * 1. prefix = common_prefix_len(ps);
280 * 2. prune something, or fill_directory
281 * 3. match_pathspec()
282 *
283 * 'prefix' at #1 may be shorter than the command's prefix and
284 * it's ok for #2 to match extra files. Those extras will be
285 * trimmed at #3.
286 *
287 * Suppose the pathspec is 'foo' and '../bar' running from
288 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
289 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
290 * user does not want XYZ/foo, only the "foo" part should be
291 * case-insensitive. We need to filter out XYZ/foo here. In
292 * other words, we do not trust the caller on comparing the
293 * prefix part when :(icase) is involved. We do exact
294 * comparison ourselves.
295 *
296 * Normally the caller (common_prefix_len() in fact) does
297 * _exact_ matching on name[-prefix+1..-1] and we do not need
298 * to check that part. Be defensive and check it anyway, in
299 * case common_prefix_len is changed, or a new caller is
300 * introduced that does not use common_prefix_len.
301 *
302 * If the penalty turns out too high when prefix is really
303 * long, maybe change it to
304 * strncmp(match, name, item->prefix - prefix)
305 */
306 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
307 strncmp(item->match, name - prefix, item->prefix))
308 return 0;
309
310 if (item->attr_match_nr && !match_attrs(name, namelen, item))
311 return 0;
312
313 /* If the match was just the prefix, we matched */
314 if (!*match)
315 return MATCHED_RECURSIVELY;
316
317 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
318 if (matchlen == namelen)
319 return MATCHED_EXACTLY;
320
321 if (match[matchlen-1] == '/' || name[matchlen] == '/')
322 return MATCHED_RECURSIVELY;
323 } else if ((flags & DO_MATCH_DIRECTORY) &&
324 match[matchlen - 1] == '/' &&
325 namelen == matchlen - 1 &&
326 !ps_strncmp(item, match, name, namelen))
327 return MATCHED_EXACTLY;
328
329 if (item->nowildcard_len < item->len &&
330 !git_fnmatch(item, match, name,
331 item->nowildcard_len - prefix))
332 return MATCHED_FNMATCH;
333
334 /* Perform checks to see if "name" is a super set of the pathspec */
335 if (flags & DO_MATCH_SUBMODULE) {
336 /* name is a literal prefix of the pathspec */
337 if ((namelen < matchlen) &&
338 (match[namelen] == '/') &&
339 !ps_strncmp(item, match, name, namelen))
340 return MATCHED_RECURSIVELY;
341
342 /* name" doesn't match up to the first wild character */
343 if (item->nowildcard_len < item->len &&
344 ps_strncmp(item, match, name,
345 item->nowildcard_len - prefix))
346 return 0;
347
348 /*
349 * Here is where we would perform a wildmatch to check if
350 * "name" can be matched as a directory (or a prefix) against
351 * the pathspec. Since wildmatch doesn't have this capability
352 * at the present we have to punt and say that it is a match,
353 * potentially returning a false positive
354 * The submodules themselves will be able to perform more
355 * accurate matching to determine if the pathspec matches.
356 */
357 return MATCHED_RECURSIVELY;
358 }
359
360 return 0;
361 }
362
363 /*
364 * Given a name and a list of pathspecs, returns the nature of the
365 * closest (i.e. most specific) match of the name to any of the
366 * pathspecs.
367 *
368 * The caller typically calls this multiple times with the same
369 * pathspec and seen[] array but with different name/namelen
370 * (e.g. entries from the index) and is interested in seeing if and
371 * how each pathspec matches all the names it calls this function
372 * with. A mark is left in the seen[] array for each pathspec element
373 * indicating the closest type of match that element achieved, so if
374 * seen[n] remains zero after multiple invocations, that means the nth
375 * pathspec did not match any names, which could indicate that the
376 * user mistyped the nth pathspec.
377 */
378 static int do_match_pathspec(const struct pathspec *ps,
379 const char *name, int namelen,
380 int prefix, char *seen,
381 unsigned flags)
382 {
383 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
384
385 GUARD_PATHSPEC(ps,
386 PATHSPEC_FROMTOP |
387 PATHSPEC_MAXDEPTH |
388 PATHSPEC_LITERAL |
389 PATHSPEC_GLOB |
390 PATHSPEC_ICASE |
391 PATHSPEC_EXCLUDE |
392 PATHSPEC_ATTR);
393
394 if (!ps->nr) {
395 if (!ps->recursive ||
396 !(ps->magic & PATHSPEC_MAXDEPTH) ||
397 ps->max_depth == -1)
398 return MATCHED_RECURSIVELY;
399
400 if (within_depth(name, namelen, 0, ps->max_depth))
401 return MATCHED_EXACTLY;
402 else
403 return 0;
404 }
405
406 name += prefix;
407 namelen -= prefix;
408
409 for (i = ps->nr - 1; i >= 0; i--) {
410 int how;
411
412 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
413 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
414 continue;
415
416 if (seen && seen[i] == MATCHED_EXACTLY)
417 continue;
418 /*
419 * Make exclude patterns optional and never report
420 * "pathspec ':(exclude)foo' matches no files"
421 */
422 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
423 seen[i] = MATCHED_FNMATCH;
424 how = match_pathspec_item(ps->items+i, prefix, name,
425 namelen, flags);
426 if (ps->recursive &&
427 (ps->magic & PATHSPEC_MAXDEPTH) &&
428 ps->max_depth != -1 &&
429 how && how != MATCHED_FNMATCH) {
430 int len = ps->items[i].len;
431 if (name[len] == '/')
432 len++;
433 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
434 how = MATCHED_EXACTLY;
435 else
436 how = 0;
437 }
438 if (how) {
439 if (retval < how)
440 retval = how;
441 if (seen && seen[i] < how)
442 seen[i] = how;
443 }
444 }
445 return retval;
446 }
447
448 int match_pathspec(const struct pathspec *ps,
449 const char *name, int namelen,
450 int prefix, char *seen, int is_dir)
451 {
452 int positive, negative;
453 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
454 positive = do_match_pathspec(ps, name, namelen,
455 prefix, seen, flags);
456 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
457 return positive;
458 negative = do_match_pathspec(ps, name, namelen,
459 prefix, seen,
460 flags | DO_MATCH_EXCLUDE);
461 return negative ? 0 : positive;
462 }
463
464 /**
465 * Check if a submodule is a superset of the pathspec
466 */
467 int submodule_path_match(const struct pathspec *ps,
468 const char *submodule_name,
469 char *seen)
470 {
471 int matched = do_match_pathspec(ps, submodule_name,
472 strlen(submodule_name),
473 0, seen,
474 DO_MATCH_DIRECTORY |
475 DO_MATCH_SUBMODULE);
476 return matched;
477 }
478
479 int report_path_error(const char *ps_matched,
480 const struct pathspec *pathspec,
481 const char *prefix)
482 {
483 /*
484 * Make sure all pathspec matched; otherwise it is an error.
485 */
486 int num, errors = 0;
487 for (num = 0; num < pathspec->nr; num++) {
488 int other, found_dup;
489
490 if (ps_matched[num])
491 continue;
492 /*
493 * The caller might have fed identical pathspec
494 * twice. Do not barf on such a mistake.
495 * FIXME: parse_pathspec should have eliminated
496 * duplicate pathspec.
497 */
498 for (found_dup = other = 0;
499 !found_dup && other < pathspec->nr;
500 other++) {
501 if (other == num || !ps_matched[other])
502 continue;
503 if (!strcmp(pathspec->items[other].original,
504 pathspec->items[num].original))
505 /*
506 * Ok, we have a match already.
507 */
508 found_dup = 1;
509 }
510 if (found_dup)
511 continue;
512
513 error("pathspec '%s' did not match any file(s) known to git.",
514 pathspec->items[num].original);
515 errors++;
516 }
517 return errors;
518 }
519
520 /*
521 * Return the length of the "simple" part of a path match limiter.
522 */
523 int simple_length(const char *match)
524 {
525 int len = -1;
526
527 for (;;) {
528 unsigned char c = *match++;
529 len++;
530 if (c == '\0' || is_glob_special(c))
531 return len;
532 }
533 }
534
535 int no_wildcard(const char *string)
536 {
537 return string[simple_length(string)] == '\0';
538 }
539
540 void parse_exclude_pattern(const char **pattern,
541 int *patternlen,
542 unsigned *flags,
543 int *nowildcardlen)
544 {
545 const char *p = *pattern;
546 size_t i, len;
547
548 *flags = 0;
549 if (*p == '!') {
550 *flags |= EXC_FLAG_NEGATIVE;
551 p++;
552 }
553 len = strlen(p);
554 if (len && p[len - 1] == '/') {
555 len--;
556 *flags |= EXC_FLAG_MUSTBEDIR;
557 }
558 for (i = 0; i < len; i++) {
559 if (p[i] == '/')
560 break;
561 }
562 if (i == len)
563 *flags |= EXC_FLAG_NODIR;
564 *nowildcardlen = simple_length(p);
565 /*
566 * we should have excluded the trailing slash from 'p' too,
567 * but that's one more allocation. Instead just make sure
568 * nowildcardlen does not exceed real patternlen
569 */
570 if (*nowildcardlen > len)
571 *nowildcardlen = len;
572 if (*p == '*' && no_wildcard(p + 1))
573 *flags |= EXC_FLAG_ENDSWITH;
574 *pattern = p;
575 *patternlen = len;
576 }
577
578 void add_exclude(const char *string, const char *base,
579 int baselen, struct exclude_list *el, int srcpos)
580 {
581 struct exclude *x;
582 int patternlen;
583 unsigned flags;
584 int nowildcardlen;
585
586 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
587 if (flags & EXC_FLAG_MUSTBEDIR) {
588 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
589 } else {
590 x = xmalloc(sizeof(*x));
591 x->pattern = string;
592 }
593 x->patternlen = patternlen;
594 x->nowildcardlen = nowildcardlen;
595 x->base = base;
596 x->baselen = baselen;
597 x->flags = flags;
598 x->srcpos = srcpos;
599 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
600 el->excludes[el->nr++] = x;
601 x->el = el;
602 }
603
604 static void *read_skip_worktree_file_from_index(const struct index_state *istate,
605 const char *path, size_t *size,
606 struct sha1_stat *sha1_stat)
607 {
608 int pos, len;
609 unsigned long sz;
610 enum object_type type;
611 void *data;
612
613 len = strlen(path);
614 pos = index_name_pos(istate, path, len);
615 if (pos < 0)
616 return NULL;
617 if (!ce_skip_worktree(istate->cache[pos]))
618 return NULL;
619 data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz);
620 if (!data || type != OBJ_BLOB) {
621 free(data);
622 return NULL;
623 }
624 *size = xsize_t(sz);
625 if (sha1_stat) {
626 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
627 hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash);
628 }
629 return data;
630 }
631
632 /*
633 * Frees memory within el which was allocated for exclude patterns and
634 * the file buffer. Does not free el itself.
635 */
636 void clear_exclude_list(struct exclude_list *el)
637 {
638 int i;
639
640 for (i = 0; i < el->nr; i++)
641 free(el->excludes[i]);
642 free(el->excludes);
643 free(el->filebuf);
644
645 memset(el, 0, sizeof(*el));
646 }
647
648 static void trim_trailing_spaces(char *buf)
649 {
650 char *p, *last_space = NULL;
651
652 for (p = buf; *p; p++)
653 switch (*p) {
654 case ' ':
655 if (!last_space)
656 last_space = p;
657 break;
658 case '\\':
659 p++;
660 if (!*p)
661 return;
662 /* fallthrough */
663 default:
664 last_space = NULL;
665 }
666
667 if (last_space)
668 *last_space = '\0';
669 }
670
671 /*
672 * Given a subdirectory name and "dir" of the current directory,
673 * search the subdir in "dir" and return it, or create a new one if it
674 * does not exist in "dir".
675 *
676 * If "name" has the trailing slash, it'll be excluded in the search.
677 */
678 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
679 struct untracked_cache_dir *dir,
680 const char *name, int len)
681 {
682 int first, last;
683 struct untracked_cache_dir *d;
684 if (!dir)
685 return NULL;
686 if (len && name[len - 1] == '/')
687 len--;
688 first = 0;
689 last = dir->dirs_nr;
690 while (last > first) {
691 int cmp, next = (last + first) >> 1;
692 d = dir->dirs[next];
693 cmp = strncmp(name, d->name, len);
694 if (!cmp && strlen(d->name) > len)
695 cmp = -1;
696 if (!cmp)
697 return d;
698 if (cmp < 0) {
699 last = next;
700 continue;
701 }
702 first = next+1;
703 }
704
705 uc->dir_created++;
706 FLEX_ALLOC_MEM(d, name, name, len);
707
708 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
709 memmove(dir->dirs + first + 1, dir->dirs + first,
710 (dir->dirs_nr - first) * sizeof(*dir->dirs));
711 dir->dirs_nr++;
712 dir->dirs[first] = d;
713 return d;
714 }
715
716 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
717 {
718 int i;
719 dir->valid = 0;
720 dir->untracked_nr = 0;
721 for (i = 0; i < dir->dirs_nr; i++)
722 do_invalidate_gitignore(dir->dirs[i]);
723 }
724
725 static void invalidate_gitignore(struct untracked_cache *uc,
726 struct untracked_cache_dir *dir)
727 {
728 uc->gitignore_invalidated++;
729 do_invalidate_gitignore(dir);
730 }
731
732 static void invalidate_directory(struct untracked_cache *uc,
733 struct untracked_cache_dir *dir)
734 {
735 int i;
736 uc->dir_invalidated++;
737 dir->valid = 0;
738 dir->untracked_nr = 0;
739 for (i = 0; i < dir->dirs_nr; i++)
740 dir->dirs[i]->recurse = 0;
741 }
742
743 /*
744 * Given a file with name "fname", read it (either from disk, or from
745 * an index if 'istate' is non-null), parse it and store the
746 * exclude rules in "el".
747 *
748 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
749 * stat data from disk (only valid if add_excludes returns zero). If
750 * ss_valid is non-zero, "ss" must contain good value as input.
751 */
752 static int add_excludes(const char *fname, const char *base, int baselen,
753 struct exclude_list *el,
754 struct index_state *istate,
755 struct sha1_stat *sha1_stat)
756 {
757 struct stat st;
758 int fd, i, lineno = 1;
759 size_t size = 0;
760 char *buf, *entry;
761
762 fd = open(fname, O_RDONLY);
763 if (fd < 0 || fstat(fd, &st) < 0) {
764 if (fd < 0)
765 warn_on_fopen_errors(fname);
766 else
767 close(fd);
768 if (!istate ||
769 (buf = read_skip_worktree_file_from_index(istate, fname, &size, sha1_stat)) == NULL)
770 return -1;
771 if (size == 0) {
772 free(buf);
773 return 0;
774 }
775 if (buf[size-1] != '\n') {
776 buf = xrealloc(buf, st_add(size, 1));
777 buf[size++] = '\n';
778 }
779 } else {
780 size = xsize_t(st.st_size);
781 if (size == 0) {
782 if (sha1_stat) {
783 fill_stat_data(&sha1_stat->stat, &st);
784 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
785 sha1_stat->valid = 1;
786 }
787 close(fd);
788 return 0;
789 }
790 buf = xmallocz(size);
791 if (read_in_full(fd, buf, size) != size) {
792 free(buf);
793 close(fd);
794 return -1;
795 }
796 buf[size++] = '\n';
797 close(fd);
798 if (sha1_stat) {
799 int pos;
800 if (sha1_stat->valid &&
801 !match_stat_data_racy(istate, &sha1_stat->stat, &st))
802 ; /* no content change, ss->sha1 still good */
803 else if (istate &&
804 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
805 !ce_stage(istate->cache[pos]) &&
806 ce_uptodate(istate->cache[pos]) &&
807 !would_convert_to_git(istate, fname))
808 hashcpy(sha1_stat->sha1,
809 istate->cache[pos]->oid.hash);
810 else
811 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
812 fill_stat_data(&sha1_stat->stat, &st);
813 sha1_stat->valid = 1;
814 }
815 }
816
817 el->filebuf = buf;
818
819 if (skip_utf8_bom(&buf, size))
820 size -= buf - el->filebuf;
821
822 entry = buf;
823
824 for (i = 0; i < size; i++) {
825 if (buf[i] == '\n') {
826 if (entry != buf + i && entry[0] != '#') {
827 buf[i - (i && buf[i-1] == '\r')] = 0;
828 trim_trailing_spaces(entry);
829 add_exclude(entry, base, baselen, el, lineno);
830 }
831 lineno++;
832 entry = buf + i + 1;
833 }
834 }
835 return 0;
836 }
837
838 int add_excludes_from_file_to_list(const char *fname, const char *base,
839 int baselen, struct exclude_list *el,
840 struct index_state *istate)
841 {
842 return add_excludes(fname, base, baselen, el, istate, NULL);
843 }
844
845 struct exclude_list *add_exclude_list(struct dir_struct *dir,
846 int group_type, const char *src)
847 {
848 struct exclude_list *el;
849 struct exclude_list_group *group;
850
851 group = &dir->exclude_list_group[group_type];
852 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
853 el = &group->el[group->nr++];
854 memset(el, 0, sizeof(*el));
855 el->src = src;
856 return el;
857 }
858
859 /*
860 * Used to set up core.excludesfile and .git/info/exclude lists.
861 */
862 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
863 struct sha1_stat *sha1_stat)
864 {
865 struct exclude_list *el;
866 /*
867 * catch setup_standard_excludes() that's called before
868 * dir->untracked is assigned. That function behaves
869 * differently when dir->untracked is non-NULL.
870 */
871 if (!dir->untracked)
872 dir->unmanaged_exclude_files++;
873 el = add_exclude_list(dir, EXC_FILE, fname);
874 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0)
875 die("cannot use %s as an exclude file", fname);
876 }
877
878 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
879 {
880 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
881 add_excludes_from_file_1(dir, fname, NULL);
882 }
883
884 int match_basename(const char *basename, int basenamelen,
885 const char *pattern, int prefix, int patternlen,
886 unsigned flags)
887 {
888 if (prefix == patternlen) {
889 if (patternlen == basenamelen &&
890 !fspathncmp(pattern, basename, basenamelen))
891 return 1;
892 } else if (flags & EXC_FLAG_ENDSWITH) {
893 /* "*literal" matching against "fooliteral" */
894 if (patternlen - 1 <= basenamelen &&
895 !fspathncmp(pattern + 1,
896 basename + basenamelen - (patternlen - 1),
897 patternlen - 1))
898 return 1;
899 } else {
900 if (fnmatch_icase_mem(pattern, patternlen,
901 basename, basenamelen,
902 0) == 0)
903 return 1;
904 }
905 return 0;
906 }
907
908 int match_pathname(const char *pathname, int pathlen,
909 const char *base, int baselen,
910 const char *pattern, int prefix, int patternlen,
911 unsigned flags)
912 {
913 const char *name;
914 int namelen;
915
916 /*
917 * match with FNM_PATHNAME; the pattern has base implicitly
918 * in front of it.
919 */
920 if (*pattern == '/') {
921 pattern++;
922 patternlen--;
923 prefix--;
924 }
925
926 /*
927 * baselen does not count the trailing slash. base[] may or
928 * may not end with a trailing slash though.
929 */
930 if (pathlen < baselen + 1 ||
931 (baselen && pathname[baselen] != '/') ||
932 fspathncmp(pathname, base, baselen))
933 return 0;
934
935 namelen = baselen ? pathlen - baselen - 1 : pathlen;
936 name = pathname + pathlen - namelen;
937
938 if (prefix) {
939 /*
940 * if the non-wildcard part is longer than the
941 * remaining pathname, surely it cannot match.
942 */
943 if (prefix > namelen)
944 return 0;
945
946 if (fspathncmp(pattern, name, prefix))
947 return 0;
948 pattern += prefix;
949 patternlen -= prefix;
950 name += prefix;
951 namelen -= prefix;
952
953 /*
954 * If the whole pattern did not have a wildcard,
955 * then our prefix match is all we need; we
956 * do not need to call fnmatch at all.
957 */
958 if (!patternlen && !namelen)
959 return 1;
960 }
961
962 return fnmatch_icase_mem(pattern, patternlen,
963 name, namelen,
964 WM_PATHNAME) == 0;
965 }
966
967 /*
968 * Scan the given exclude list in reverse to see whether pathname
969 * should be ignored. The first match (i.e. the last on the list), if
970 * any, determines the fate. Returns the exclude_list element which
971 * matched, or NULL for undecided.
972 */
973 static struct exclude *last_exclude_matching_from_list(const char *pathname,
974 int pathlen,
975 const char *basename,
976 int *dtype,
977 struct exclude_list *el,
978 struct index_state *istate)
979 {
980 struct exclude *exc = NULL; /* undecided */
981 int i;
982
983 if (!el->nr)
984 return NULL; /* undefined */
985
986 for (i = el->nr - 1; 0 <= i; i--) {
987 struct exclude *x = el->excludes[i];
988 const char *exclude = x->pattern;
989 int prefix = x->nowildcardlen;
990
991 if (x->flags & EXC_FLAG_MUSTBEDIR) {
992 if (*dtype == DT_UNKNOWN)
993 *dtype = get_dtype(NULL, istate, pathname, pathlen);
994 if (*dtype != DT_DIR)
995 continue;
996 }
997
998 if (x->flags & EXC_FLAG_NODIR) {
999 if (match_basename(basename,
1000 pathlen - (basename - pathname),
1001 exclude, prefix, x->patternlen,
1002 x->flags)) {
1003 exc = x;
1004 break;
1005 }
1006 continue;
1007 }
1008
1009 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1010 if (match_pathname(pathname, pathlen,
1011 x->base, x->baselen ? x->baselen - 1 : 0,
1012 exclude, prefix, x->patternlen, x->flags)) {
1013 exc = x;
1014 break;
1015 }
1016 }
1017 return exc;
1018 }
1019
1020 /*
1021 * Scan the list and let the last match determine the fate.
1022 * Return 1 for exclude, 0 for include and -1 for undecided.
1023 */
1024 int is_excluded_from_list(const char *pathname,
1025 int pathlen, const char *basename, int *dtype,
1026 struct exclude_list *el, struct index_state *istate)
1027 {
1028 struct exclude *exclude;
1029 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1030 dtype, el, istate);
1031 if (exclude)
1032 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1033 return -1; /* undecided */
1034 }
1035
1036 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1037 struct index_state *istate,
1038 const char *pathname, int pathlen, const char *basename,
1039 int *dtype_p)
1040 {
1041 int i, j;
1042 struct exclude_list_group *group;
1043 struct exclude *exclude;
1044 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1045 group = &dir->exclude_list_group[i];
1046 for (j = group->nr - 1; j >= 0; j--) {
1047 exclude = last_exclude_matching_from_list(
1048 pathname, pathlen, basename, dtype_p,
1049 &group->el[j], istate);
1050 if (exclude)
1051 return exclude;
1052 }
1053 }
1054 return NULL;
1055 }
1056
1057 /*
1058 * Loads the per-directory exclude list for the substring of base
1059 * which has a char length of baselen.
1060 */
1061 static void prep_exclude(struct dir_struct *dir,
1062 struct index_state *istate,
1063 const char *base, int baselen)
1064 {
1065 struct exclude_list_group *group;
1066 struct exclude_list *el;
1067 struct exclude_stack *stk = NULL;
1068 struct untracked_cache_dir *untracked;
1069 int current;
1070
1071 group = &dir->exclude_list_group[EXC_DIRS];
1072
1073 /*
1074 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1075 * which originate from directories not in the prefix of the
1076 * path being checked.
1077 */
1078 while ((stk = dir->exclude_stack) != NULL) {
1079 if (stk->baselen <= baselen &&
1080 !strncmp(dir->basebuf.buf, base, stk->baselen))
1081 break;
1082 el = &group->el[dir->exclude_stack->exclude_ix];
1083 dir->exclude_stack = stk->prev;
1084 dir->exclude = NULL;
1085 free((char *)el->src); /* see strbuf_detach() below */
1086 clear_exclude_list(el);
1087 free(stk);
1088 group->nr--;
1089 }
1090
1091 /* Skip traversing into sub directories if the parent is excluded */
1092 if (dir->exclude)
1093 return;
1094
1095 /*
1096 * Lazy initialization. All call sites currently just
1097 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1098 * them seems lots of work for little benefit.
1099 */
1100 if (!dir->basebuf.buf)
1101 strbuf_init(&dir->basebuf, PATH_MAX);
1102
1103 /* Read from the parent directories and push them down. */
1104 current = stk ? stk->baselen : -1;
1105 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1106 if (dir->untracked)
1107 untracked = stk ? stk->ucd : dir->untracked->root;
1108 else
1109 untracked = NULL;
1110
1111 while (current < baselen) {
1112 const char *cp;
1113 struct sha1_stat sha1_stat;
1114
1115 stk = xcalloc(1, sizeof(*stk));
1116 if (current < 0) {
1117 cp = base;
1118 current = 0;
1119 } else {
1120 cp = strchr(base + current + 1, '/');
1121 if (!cp)
1122 die("oops in prep_exclude");
1123 cp++;
1124 untracked =
1125 lookup_untracked(dir->untracked, untracked,
1126 base + current,
1127 cp - base - current);
1128 }
1129 stk->prev = dir->exclude_stack;
1130 stk->baselen = cp - base;
1131 stk->exclude_ix = group->nr;
1132 stk->ucd = untracked;
1133 el = add_exclude_list(dir, EXC_DIRS, NULL);
1134 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1135 assert(stk->baselen == dir->basebuf.len);
1136
1137 /* Abort if the directory is excluded */
1138 if (stk->baselen) {
1139 int dt = DT_DIR;
1140 dir->basebuf.buf[stk->baselen - 1] = 0;
1141 dir->exclude = last_exclude_matching_from_lists(dir,
1142 istate,
1143 dir->basebuf.buf, stk->baselen - 1,
1144 dir->basebuf.buf + current, &dt);
1145 dir->basebuf.buf[stk->baselen - 1] = '/';
1146 if (dir->exclude &&
1147 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1148 dir->exclude = NULL;
1149 if (dir->exclude) {
1150 dir->exclude_stack = stk;
1151 return;
1152 }
1153 }
1154
1155 /* Try to read per-directory file */
1156 hashclr(sha1_stat.sha1);
1157 sha1_stat.valid = 0;
1158 if (dir->exclude_per_dir &&
1159 /*
1160 * If we know that no files have been added in
1161 * this directory (i.e. valid_cached_dir() has
1162 * been executed and set untracked->valid) ..
1163 */
1164 (!untracked || !untracked->valid ||
1165 /*
1166 * .. and .gitignore does not exist before
1167 * (i.e. null exclude_sha1). Then we can skip
1168 * loading .gitignore, which would result in
1169 * ENOENT anyway.
1170 */
1171 !is_null_sha1(untracked->exclude_sha1))) {
1172 /*
1173 * dir->basebuf gets reused by the traversal, but we
1174 * need fname to remain unchanged to ensure the src
1175 * member of each struct exclude correctly
1176 * back-references its source file. Other invocations
1177 * of add_exclude_list provide stable strings, so we
1178 * strbuf_detach() and free() here in the caller.
1179 */
1180 struct strbuf sb = STRBUF_INIT;
1181 strbuf_addbuf(&sb, &dir->basebuf);
1182 strbuf_addstr(&sb, dir->exclude_per_dir);
1183 el->src = strbuf_detach(&sb, NULL);
1184 add_excludes(el->src, el->src, stk->baselen, el, istate,
1185 untracked ? &sha1_stat : NULL);
1186 }
1187 /*
1188 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1189 * will first be called in valid_cached_dir() then maybe many
1190 * times more in last_exclude_matching(). When the cache is
1191 * used, last_exclude_matching() will not be called and
1192 * reading .gitignore content will be a waste.
1193 *
1194 * So when it's called by valid_cached_dir() and we can get
1195 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1196 * modified on work tree), we could delay reading the
1197 * .gitignore content until we absolutely need it in
1198 * last_exclude_matching(). Be careful about ignore rule
1199 * order, though, if you do that.
1200 */
1201 if (untracked &&
1202 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1203 invalidate_gitignore(dir->untracked, untracked);
1204 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1205 }
1206 dir->exclude_stack = stk;
1207 current = stk->baselen;
1208 }
1209 strbuf_setlen(&dir->basebuf, baselen);
1210 }
1211
1212 /*
1213 * Loads the exclude lists for the directory containing pathname, then
1214 * scans all exclude lists to determine whether pathname is excluded.
1215 * Returns the exclude_list element which matched, or NULL for
1216 * undecided.
1217 */
1218 struct exclude *last_exclude_matching(struct dir_struct *dir,
1219 struct index_state *istate,
1220 const char *pathname,
1221 int *dtype_p)
1222 {
1223 int pathlen = strlen(pathname);
1224 const char *basename = strrchr(pathname, '/');
1225 basename = (basename) ? basename+1 : pathname;
1226
1227 prep_exclude(dir, istate, pathname, basename-pathname);
1228
1229 if (dir->exclude)
1230 return dir->exclude;
1231
1232 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1233 basename, dtype_p);
1234 }
1235
1236 /*
1237 * Loads the exclude lists for the directory containing pathname, then
1238 * scans all exclude lists to determine whether pathname is excluded.
1239 * Returns 1 if true, otherwise 0.
1240 */
1241 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1242 const char *pathname, int *dtype_p)
1243 {
1244 struct exclude *exclude =
1245 last_exclude_matching(dir, istate, pathname, dtype_p);
1246 if (exclude)
1247 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1248 return 0;
1249 }
1250
1251 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1252 {
1253 struct dir_entry *ent;
1254
1255 FLEX_ALLOC_MEM(ent, name, pathname, len);
1256 ent->len = len;
1257 return ent;
1258 }
1259
1260 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1261 struct index_state *istate,
1262 const char *pathname, int len)
1263 {
1264 if (index_file_exists(istate, pathname, len, ignore_case))
1265 return NULL;
1266
1267 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1268 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1269 }
1270
1271 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1272 struct index_state *istate,
1273 const char *pathname, int len)
1274 {
1275 if (!index_name_is_other(istate, pathname, len))
1276 return NULL;
1277
1278 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1279 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1280 }
1281
1282 enum exist_status {
1283 index_nonexistent = 0,
1284 index_directory,
1285 index_gitdir
1286 };
1287
1288 /*
1289 * Do not use the alphabetically sorted index to look up
1290 * the directory name; instead, use the case insensitive
1291 * directory hash.
1292 */
1293 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1294 const char *dirname, int len)
1295 {
1296 struct cache_entry *ce;
1297
1298 if (index_dir_exists(istate, dirname, len))
1299 return index_directory;
1300
1301 ce = index_file_exists(istate, dirname, len, ignore_case);
1302 if (ce && S_ISGITLINK(ce->ce_mode))
1303 return index_gitdir;
1304
1305 return index_nonexistent;
1306 }
1307
1308 /*
1309 * The index sorts alphabetically by entry name, which
1310 * means that a gitlink sorts as '\0' at the end, while
1311 * a directory (which is defined not as an entry, but as
1312 * the files it contains) will sort with the '/' at the
1313 * end.
1314 */
1315 static enum exist_status directory_exists_in_index(struct index_state *istate,
1316 const char *dirname, int len)
1317 {
1318 int pos;
1319
1320 if (ignore_case)
1321 return directory_exists_in_index_icase(istate, dirname, len);
1322
1323 pos = index_name_pos(istate, dirname, len);
1324 if (pos < 0)
1325 pos = -pos-1;
1326 while (pos < istate->cache_nr) {
1327 const struct cache_entry *ce = istate->cache[pos++];
1328 unsigned char endchar;
1329
1330 if (strncmp(ce->name, dirname, len))
1331 break;
1332 endchar = ce->name[len];
1333 if (endchar > '/')
1334 break;
1335 if (endchar == '/')
1336 return index_directory;
1337 if (!endchar && S_ISGITLINK(ce->ce_mode))
1338 return index_gitdir;
1339 }
1340 return index_nonexistent;
1341 }
1342
1343 /*
1344 * When we find a directory when traversing the filesystem, we
1345 * have three distinct cases:
1346 *
1347 * - ignore it
1348 * - see it as a directory
1349 * - recurse into it
1350 *
1351 * and which one we choose depends on a combination of existing
1352 * git index contents and the flags passed into the directory
1353 * traversal routine.
1354 *
1355 * Case 1: If we *already* have entries in the index under that
1356 * directory name, we always recurse into the directory to see
1357 * all the files.
1358 *
1359 * Case 2: If we *already* have that directory name as a gitlink,
1360 * we always continue to see it as a gitlink, regardless of whether
1361 * there is an actual git directory there or not (it might not
1362 * be checked out as a subproject!)
1363 *
1364 * Case 3: if we didn't have it in the index previously, we
1365 * have a few sub-cases:
1366 *
1367 * (a) if "show_other_directories" is true, we show it as
1368 * just a directory, unless "hide_empty_directories" is
1369 * also true, in which case we need to check if it contains any
1370 * untracked and / or ignored files.
1371 * (b) if it looks like a git directory, and we don't have
1372 * 'no_gitlinks' set we treat it as a gitlink, and show it
1373 * as a directory.
1374 * (c) otherwise, we recurse into it.
1375 */
1376 static enum path_treatment treat_directory(struct dir_struct *dir,
1377 struct index_state *istate,
1378 struct untracked_cache_dir *untracked,
1379 const char *dirname, int len, int baselen, int exclude,
1380 const struct pathspec *pathspec)
1381 {
1382 /* The "len-1" is to strip the final '/' */
1383 switch (directory_exists_in_index(istate, dirname, len-1)) {
1384 case index_directory:
1385 return path_recurse;
1386
1387 case index_gitdir:
1388 return path_none;
1389
1390 case index_nonexistent:
1391 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1392 break;
1393 if (exclude &&
1394 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1395 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1396
1397 /*
1398 * This is an excluded directory and we are
1399 * showing ignored paths that match an exclude
1400 * pattern. (e.g. show directory as ignored
1401 * only if it matches an exclude pattern).
1402 * This path will either be 'path_excluded`
1403 * (if we are showing empty directories or if
1404 * the directory is not empty), or will be
1405 * 'path_none' (empty directory, and we are
1406 * not showing empty directories).
1407 */
1408 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1409 return path_excluded;
1410
1411 if (read_directory_recursive(dir, istate, dirname, len,
1412 untracked, 1, 1, pathspec) == path_excluded)
1413 return path_excluded;
1414
1415 return path_none;
1416 }
1417 if (!(dir->flags & DIR_NO_GITLINKS)) {
1418 struct object_id oid;
1419 if (resolve_gitlink_ref(dirname, "HEAD", &oid) == 0)
1420 return exclude ? path_excluded : path_untracked;
1421 }
1422 return path_recurse;
1423 }
1424
1425 /* This is the "show_other_directories" case */
1426
1427 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1428 return exclude ? path_excluded : path_untracked;
1429
1430 untracked = lookup_untracked(dir->untracked, untracked,
1431 dirname + baselen, len - baselen);
1432
1433 /*
1434 * If this is an excluded directory, then we only need to check if
1435 * the directory contains any files.
1436 */
1437 return read_directory_recursive(dir, istate, dirname, len,
1438 untracked, 1, exclude, pathspec);
1439 }
1440
1441 /*
1442 * This is an inexact early pruning of any recursive directory
1443 * reading - if the path cannot possibly be in the pathspec,
1444 * return true, and we'll skip it early.
1445 */
1446 static int simplify_away(const char *path, int pathlen,
1447 const struct pathspec *pathspec)
1448 {
1449 int i;
1450
1451 if (!pathspec || !pathspec->nr)
1452 return 0;
1453
1454 GUARD_PATHSPEC(pathspec,
1455 PATHSPEC_FROMTOP |
1456 PATHSPEC_MAXDEPTH |
1457 PATHSPEC_LITERAL |
1458 PATHSPEC_GLOB |
1459 PATHSPEC_ICASE |
1460 PATHSPEC_EXCLUDE |
1461 PATHSPEC_ATTR);
1462
1463 for (i = 0; i < pathspec->nr; i++) {
1464 const struct pathspec_item *item = &pathspec->items[i];
1465 int len = item->nowildcard_len;
1466
1467 if (len > pathlen)
1468 len = pathlen;
1469 if (!ps_strncmp(item, item->match, path, len))
1470 return 0;
1471 }
1472
1473 return 1;
1474 }
1475
1476 /*
1477 * This function tells us whether an excluded path matches a
1478 * list of "interesting" pathspecs. That is, whether a path matched
1479 * by any of the pathspecs could possibly be ignored by excluding
1480 * the specified path. This can happen if:
1481 *
1482 * 1. the path is mentioned explicitly in the pathspec
1483 *
1484 * 2. the path is a directory prefix of some element in the
1485 * pathspec
1486 */
1487 static int exclude_matches_pathspec(const char *path, int pathlen,
1488 const struct pathspec *pathspec)
1489 {
1490 int i;
1491
1492 if (!pathspec || !pathspec->nr)
1493 return 0;
1494
1495 GUARD_PATHSPEC(pathspec,
1496 PATHSPEC_FROMTOP |
1497 PATHSPEC_MAXDEPTH |
1498 PATHSPEC_LITERAL |
1499 PATHSPEC_GLOB |
1500 PATHSPEC_ICASE |
1501 PATHSPEC_EXCLUDE);
1502
1503 for (i = 0; i < pathspec->nr; i++) {
1504 const struct pathspec_item *item = &pathspec->items[i];
1505 int len = item->nowildcard_len;
1506
1507 if (len == pathlen &&
1508 !ps_strncmp(item, item->match, path, pathlen))
1509 return 1;
1510 if (len > pathlen &&
1511 item->match[pathlen] == '/' &&
1512 !ps_strncmp(item, item->match, path, pathlen))
1513 return 1;
1514 }
1515 return 0;
1516 }
1517
1518 static int get_index_dtype(struct index_state *istate,
1519 const char *path, int len)
1520 {
1521 int pos;
1522 const struct cache_entry *ce;
1523
1524 ce = index_file_exists(istate, path, len, 0);
1525 if (ce) {
1526 if (!ce_uptodate(ce))
1527 return DT_UNKNOWN;
1528 if (S_ISGITLINK(ce->ce_mode))
1529 return DT_DIR;
1530 /*
1531 * Nobody actually cares about the
1532 * difference between DT_LNK and DT_REG
1533 */
1534 return DT_REG;
1535 }
1536
1537 /* Try to look it up as a directory */
1538 pos = index_name_pos(istate, path, len);
1539 if (pos >= 0)
1540 return DT_UNKNOWN;
1541 pos = -pos-1;
1542 while (pos < istate->cache_nr) {
1543 ce = istate->cache[pos++];
1544 if (strncmp(ce->name, path, len))
1545 break;
1546 if (ce->name[len] > '/')
1547 break;
1548 if (ce->name[len] < '/')
1549 continue;
1550 if (!ce_uptodate(ce))
1551 break; /* continue? */
1552 return DT_DIR;
1553 }
1554 return DT_UNKNOWN;
1555 }
1556
1557 static int get_dtype(struct dirent *de, struct index_state *istate,
1558 const char *path, int len)
1559 {
1560 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1561 struct stat st;
1562
1563 if (dtype != DT_UNKNOWN)
1564 return dtype;
1565 dtype = get_index_dtype(istate, path, len);
1566 if (dtype != DT_UNKNOWN)
1567 return dtype;
1568 if (lstat(path, &st))
1569 return dtype;
1570 if (S_ISREG(st.st_mode))
1571 return DT_REG;
1572 if (S_ISDIR(st.st_mode))
1573 return DT_DIR;
1574 if (S_ISLNK(st.st_mode))
1575 return DT_LNK;
1576 return dtype;
1577 }
1578
1579 static enum path_treatment treat_one_path(struct dir_struct *dir,
1580 struct untracked_cache_dir *untracked,
1581 struct index_state *istate,
1582 struct strbuf *path,
1583 int baselen,
1584 const struct pathspec *pathspec,
1585 int dtype, struct dirent *de)
1586 {
1587 int exclude;
1588 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1589 enum path_treatment path_treatment;
1590
1591 if (dtype == DT_UNKNOWN)
1592 dtype = get_dtype(de, istate, path->buf, path->len);
1593
1594 /* Always exclude indexed files */
1595 if (dtype != DT_DIR && has_path_in_index)
1596 return path_none;
1597
1598 /*
1599 * When we are looking at a directory P in the working tree,
1600 * there are three cases:
1601 *
1602 * (1) P exists in the index. Everything inside the directory P in
1603 * the working tree needs to go when P is checked out from the
1604 * index.
1605 *
1606 * (2) P does not exist in the index, but there is P/Q in the index.
1607 * We know P will stay a directory when we check out the contents
1608 * of the index, but we do not know yet if there is a directory
1609 * P/Q in the working tree to be killed, so we need to recurse.
1610 *
1611 * (3) P does not exist in the index, and there is no P/Q in the index
1612 * to require P to be a directory, either. Only in this case, we
1613 * know that everything inside P will not be killed without
1614 * recursing.
1615 */
1616 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1617 (dtype == DT_DIR) &&
1618 !has_path_in_index &&
1619 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1620 return path_none;
1621
1622 exclude = is_excluded(dir, istate, path->buf, &dtype);
1623
1624 /*
1625 * Excluded? If we don't explicitly want to show
1626 * ignored files, ignore it
1627 */
1628 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1629 return path_excluded;
1630
1631 switch (dtype) {
1632 default:
1633 return path_none;
1634 case DT_DIR:
1635 strbuf_addch(path, '/');
1636 path_treatment = treat_directory(dir, istate, untracked,
1637 path->buf, path->len,
1638 baselen, exclude, pathspec);
1639 /*
1640 * If 1) we only want to return directories that
1641 * match an exclude pattern and 2) this directory does
1642 * not match an exclude pattern but all of its
1643 * contents are excluded, then indicate that we should
1644 * recurse into this directory (instead of marking the
1645 * directory itself as an ignored path).
1646 */
1647 if (!exclude &&
1648 path_treatment == path_excluded &&
1649 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1650 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1651 return path_recurse;
1652 return path_treatment;
1653 case DT_REG:
1654 case DT_LNK:
1655 return exclude ? path_excluded : path_untracked;
1656 }
1657 }
1658
1659 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1660 struct untracked_cache_dir *untracked,
1661 struct cached_dir *cdir,
1662 struct index_state *istate,
1663 struct strbuf *path,
1664 int baselen,
1665 const struct pathspec *pathspec)
1666 {
1667 strbuf_setlen(path, baselen);
1668 if (!cdir->ucd) {
1669 strbuf_addstr(path, cdir->file);
1670 return path_untracked;
1671 }
1672 strbuf_addstr(path, cdir->ucd->name);
1673 /* treat_one_path() does this before it calls treat_directory() */
1674 strbuf_complete(path, '/');
1675 if (cdir->ucd->check_only)
1676 /*
1677 * check_only is set as a result of treat_directory() getting
1678 * to its bottom. Verify again the same set of directories
1679 * with check_only set.
1680 */
1681 return read_directory_recursive(dir, istate, path->buf, path->len,
1682 cdir->ucd, 1, 0, pathspec);
1683 /*
1684 * We get path_recurse in the first run when
1685 * directory_exists_in_index() returns index_nonexistent. We
1686 * are sure that new changes in the index does not impact the
1687 * outcome. Return now.
1688 */
1689 return path_recurse;
1690 }
1691
1692 static enum path_treatment treat_path(struct dir_struct *dir,
1693 struct untracked_cache_dir *untracked,
1694 struct cached_dir *cdir,
1695 struct index_state *istate,
1696 struct strbuf *path,
1697 int baselen,
1698 const struct pathspec *pathspec)
1699 {
1700 int dtype;
1701 struct dirent *de = cdir->de;
1702
1703 if (!de)
1704 return treat_path_fast(dir, untracked, cdir, istate, path,
1705 baselen, pathspec);
1706 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1707 return path_none;
1708 strbuf_setlen(path, baselen);
1709 strbuf_addstr(path, de->d_name);
1710 if (simplify_away(path->buf, path->len, pathspec))
1711 return path_none;
1712
1713 dtype = DTYPE(de);
1714 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1715 }
1716
1717 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1718 {
1719 if (!dir)
1720 return;
1721 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1722 dir->untracked_alloc);
1723 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1724 }
1725
1726 static int valid_cached_dir(struct dir_struct *dir,
1727 struct untracked_cache_dir *untracked,
1728 struct index_state *istate,
1729 struct strbuf *path,
1730 int check_only)
1731 {
1732 struct stat st;
1733
1734 if (!untracked)
1735 return 0;
1736
1737 /*
1738 * With fsmonitor, we can trust the untracked cache's valid field.
1739 */
1740 refresh_fsmonitor(istate);
1741 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1742 if (stat(path->len ? path->buf : ".", &st)) {
1743 invalidate_directory(dir->untracked, untracked);
1744 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1745 return 0;
1746 }
1747 if (!untracked->valid ||
1748 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1749 if (untracked->valid)
1750 invalidate_directory(dir->untracked, untracked);
1751 fill_stat_data(&untracked->stat_data, &st);
1752 return 0;
1753 }
1754 }
1755
1756 if (untracked->check_only != !!check_only) {
1757 invalidate_directory(dir->untracked, untracked);
1758 return 0;
1759 }
1760
1761 /*
1762 * prep_exclude will be called eventually on this directory,
1763 * but it's called much later in last_exclude_matching(). We
1764 * need it now to determine the validity of the cache for this
1765 * path. The next calls will be nearly no-op, the way
1766 * prep_exclude() is designed.
1767 */
1768 if (path->len && path->buf[path->len - 1] != '/') {
1769 strbuf_addch(path, '/');
1770 prep_exclude(dir, istate, path->buf, path->len);
1771 strbuf_setlen(path, path->len - 1);
1772 } else
1773 prep_exclude(dir, istate, path->buf, path->len);
1774
1775 /* hopefully prep_exclude() haven't invalidated this entry... */
1776 return untracked->valid;
1777 }
1778
1779 static int open_cached_dir(struct cached_dir *cdir,
1780 struct dir_struct *dir,
1781 struct untracked_cache_dir *untracked,
1782 struct index_state *istate,
1783 struct strbuf *path,
1784 int check_only)
1785 {
1786 memset(cdir, 0, sizeof(*cdir));
1787 cdir->untracked = untracked;
1788 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1789 return 0;
1790 cdir->fdir = opendir(path->len ? path->buf : ".");
1791 if (dir->untracked)
1792 dir->untracked->dir_opened++;
1793 if (!cdir->fdir)
1794 return -1;
1795 return 0;
1796 }
1797
1798 static int read_cached_dir(struct cached_dir *cdir)
1799 {
1800 if (cdir->fdir) {
1801 cdir->de = readdir(cdir->fdir);
1802 if (!cdir->de)
1803 return -1;
1804 return 0;
1805 }
1806 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1807 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1808 if (!d->recurse) {
1809 cdir->nr_dirs++;
1810 continue;
1811 }
1812 cdir->ucd = d;
1813 cdir->nr_dirs++;
1814 return 0;
1815 }
1816 cdir->ucd = NULL;
1817 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1818 struct untracked_cache_dir *d = cdir->untracked;
1819 cdir->file = d->untracked[cdir->nr_files++];
1820 return 0;
1821 }
1822 return -1;
1823 }
1824
1825 static void close_cached_dir(struct cached_dir *cdir)
1826 {
1827 if (cdir->fdir)
1828 closedir(cdir->fdir);
1829 /*
1830 * We have gone through this directory and found no untracked
1831 * entries. Mark it valid.
1832 */
1833 if (cdir->untracked) {
1834 cdir->untracked->valid = 1;
1835 cdir->untracked->recurse = 1;
1836 }
1837 }
1838
1839 /*
1840 * Read a directory tree. We currently ignore anything but
1841 * directories, regular files and symlinks. That's because git
1842 * doesn't handle them at all yet. Maybe that will change some
1843 * day.
1844 *
1845 * Also, we ignore the name ".git" (even if it is not a directory).
1846 * That likely will not change.
1847 *
1848 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1849 * to signal that a file was found. This is the least significant value that
1850 * indicates that a file was encountered that does not depend on the order of
1851 * whether an untracked or exluded path was encountered first.
1852 *
1853 * Returns the most significant path_treatment value encountered in the scan.
1854 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1855 * significant path_treatment value that will be returned.
1856 */
1857
1858 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1859 struct index_state *istate, const char *base, int baselen,
1860 struct untracked_cache_dir *untracked, int check_only,
1861 int stop_at_first_file, const struct pathspec *pathspec)
1862 {
1863 struct cached_dir cdir;
1864 enum path_treatment state, subdir_state, dir_state = path_none;
1865 struct strbuf path = STRBUF_INIT;
1866
1867 strbuf_add(&path, base, baselen);
1868
1869 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1870 goto out;
1871
1872 if (untracked)
1873 untracked->check_only = !!check_only;
1874
1875 while (!read_cached_dir(&cdir)) {
1876 /* check how the file or directory should be treated */
1877 state = treat_path(dir, untracked, &cdir, istate, &path,
1878 baselen, pathspec);
1879
1880 if (state > dir_state)
1881 dir_state = state;
1882
1883 /* recurse into subdir if instructed by treat_path */
1884 if ((state == path_recurse) ||
1885 ((state == path_untracked) &&
1886 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1887 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1888 struct untracked_cache_dir *ud;
1889 ud = lookup_untracked(dir->untracked, untracked,
1890 path.buf + baselen,
1891 path.len - baselen);
1892 subdir_state =
1893 read_directory_recursive(dir, istate, path.buf,
1894 path.len, ud,
1895 check_only, stop_at_first_file, pathspec);
1896 if (subdir_state > dir_state)
1897 dir_state = subdir_state;
1898 }
1899
1900 if (check_only) {
1901 if (stop_at_first_file) {
1902 /*
1903 * If stopping at first file, then
1904 * signal that a file was found by
1905 * returning `path_excluded`. This is
1906 * to return a consistent value
1907 * regardless of whether an ignored or
1908 * excluded file happened to be
1909 * encountered 1st.
1910 *
1911 * In current usage, the
1912 * `stop_at_first_file` is passed when
1913 * an ancestor directory has matched
1914 * an exclude pattern, so any found
1915 * files will be excluded.
1916 */
1917 if (dir_state >= path_excluded) {
1918 dir_state = path_excluded;
1919 break;
1920 }
1921 }
1922
1923 /* abort early if maximum state has been reached */
1924 if (dir_state == path_untracked) {
1925 if (cdir.fdir)
1926 add_untracked(untracked, path.buf + baselen);
1927 break;
1928 }
1929 /* skip the dir_add_* part */
1930 continue;
1931 }
1932
1933 /* add the path to the appropriate result list */
1934 switch (state) {
1935 case path_excluded:
1936 if (dir->flags & DIR_SHOW_IGNORED)
1937 dir_add_name(dir, istate, path.buf, path.len);
1938 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1939 ((dir->flags & DIR_COLLECT_IGNORED) &&
1940 exclude_matches_pathspec(path.buf, path.len,
1941 pathspec)))
1942 dir_add_ignored(dir, istate, path.buf, path.len);
1943 break;
1944
1945 case path_untracked:
1946 if (dir->flags & DIR_SHOW_IGNORED)
1947 break;
1948 dir_add_name(dir, istate, path.buf, path.len);
1949 if (cdir.fdir)
1950 add_untracked(untracked, path.buf + baselen);
1951 break;
1952
1953 default:
1954 break;
1955 }
1956 }
1957 close_cached_dir(&cdir);
1958 out:
1959 strbuf_release(&path);
1960
1961 return dir_state;
1962 }
1963
1964 int cmp_dir_entry(const void *p1, const void *p2)
1965 {
1966 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1967 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1968
1969 return name_compare(e1->name, e1->len, e2->name, e2->len);
1970 }
1971
1972 /* check if *out lexically strictly contains *in */
1973 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
1974 {
1975 return (out->len < in->len) &&
1976 (out->name[out->len - 1] == '/') &&
1977 !memcmp(out->name, in->name, out->len);
1978 }
1979
1980 static int treat_leading_path(struct dir_struct *dir,
1981 struct index_state *istate,
1982 const char *path, int len,
1983 const struct pathspec *pathspec)
1984 {
1985 struct strbuf sb = STRBUF_INIT;
1986 int baselen, rc = 0;
1987 const char *cp;
1988 int old_flags = dir->flags;
1989
1990 while (len && path[len - 1] == '/')
1991 len--;
1992 if (!len)
1993 return 1;
1994 baselen = 0;
1995 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1996 while (1) {
1997 cp = path + baselen + !!baselen;
1998 cp = memchr(cp, '/', path + len - cp);
1999 if (!cp)
2000 baselen = len;
2001 else
2002 baselen = cp - path;
2003 strbuf_setlen(&sb, 0);
2004 strbuf_add(&sb, path, baselen);
2005 if (!is_directory(sb.buf))
2006 break;
2007 if (simplify_away(sb.buf, sb.len, pathspec))
2008 break;
2009 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2010 DT_DIR, NULL) == path_none)
2011 break; /* do not recurse into it */
2012 if (len <= baselen) {
2013 rc = 1;
2014 break; /* finished checking */
2015 }
2016 }
2017 strbuf_release(&sb);
2018 dir->flags = old_flags;
2019 return rc;
2020 }
2021
2022 static const char *get_ident_string(void)
2023 {
2024 static struct strbuf sb = STRBUF_INIT;
2025 struct utsname uts;
2026
2027 if (sb.len)
2028 return sb.buf;
2029 if (uname(&uts) < 0)
2030 die_errno(_("failed to get kernel name and information"));
2031 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2032 uts.sysname);
2033 return sb.buf;
2034 }
2035
2036 static int ident_in_untracked(const struct untracked_cache *uc)
2037 {
2038 /*
2039 * Previous git versions may have saved many NUL separated
2040 * strings in the "ident" field, but it is insane to manage
2041 * many locations, so just take care of the first one.
2042 */
2043
2044 return !strcmp(uc->ident.buf, get_ident_string());
2045 }
2046
2047 static void set_untracked_ident(struct untracked_cache *uc)
2048 {
2049 strbuf_reset(&uc->ident);
2050 strbuf_addstr(&uc->ident, get_ident_string());
2051
2052 /*
2053 * This strbuf used to contain a list of NUL separated
2054 * strings, so save NUL too for backward compatibility.
2055 */
2056 strbuf_addch(&uc->ident, 0);
2057 }
2058
2059 static void new_untracked_cache(struct index_state *istate)
2060 {
2061 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2062 strbuf_init(&uc->ident, 100);
2063 uc->exclude_per_dir = ".gitignore";
2064 /* should be the same flags used by git-status */
2065 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2066 set_untracked_ident(uc);
2067 istate->untracked = uc;
2068 istate->cache_changed |= UNTRACKED_CHANGED;
2069 }
2070
2071 void add_untracked_cache(struct index_state *istate)
2072 {
2073 if (!istate->untracked) {
2074 new_untracked_cache(istate);
2075 } else {
2076 if (!ident_in_untracked(istate->untracked)) {
2077 free_untracked_cache(istate->untracked);
2078 new_untracked_cache(istate);
2079 }
2080 }
2081 }
2082
2083 void remove_untracked_cache(struct index_state *istate)
2084 {
2085 if (istate->untracked) {
2086 free_untracked_cache(istate->untracked);
2087 istate->untracked = NULL;
2088 istate->cache_changed |= UNTRACKED_CHANGED;
2089 }
2090 }
2091
2092 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2093 int base_len,
2094 const struct pathspec *pathspec)
2095 {
2096 struct untracked_cache_dir *root;
2097
2098 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
2099 return NULL;
2100
2101 /*
2102 * We only support $GIT_DIR/info/exclude and core.excludesfile
2103 * as the global ignore rule files. Any other additions
2104 * (e.g. from command line) invalidate the cache. This
2105 * condition also catches running setup_standard_excludes()
2106 * before setting dir->untracked!
2107 */
2108 if (dir->unmanaged_exclude_files)
2109 return NULL;
2110
2111 /*
2112 * Optimize for the main use case only: whole-tree git
2113 * status. More work involved in treat_leading_path() if we
2114 * use cache on just a subset of the worktree. pathspec
2115 * support could make the matter even worse.
2116 */
2117 if (base_len || (pathspec && pathspec->nr))
2118 return NULL;
2119
2120 /* Different set of flags may produce different results */
2121 if (dir->flags != dir->untracked->dir_flags ||
2122 /*
2123 * See treat_directory(), case index_nonexistent. Without
2124 * this flag, we may need to also cache .git file content
2125 * for the resolve_gitlink_ref() call, which we don't.
2126 */
2127 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2128 /* We don't support collecting ignore files */
2129 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2130 DIR_COLLECT_IGNORED)))
2131 return NULL;
2132
2133 /*
2134 * If we use .gitignore in the cache and now you change it to
2135 * .gitexclude, everything will go wrong.
2136 */
2137 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2138 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2139 return NULL;
2140
2141 /*
2142 * EXC_CMDL is not considered in the cache. If people set it,
2143 * skip the cache.
2144 */
2145 if (dir->exclude_list_group[EXC_CMDL].nr)
2146 return NULL;
2147
2148 if (!ident_in_untracked(dir->untracked)) {
2149 warning(_("Untracked cache is disabled on this system or location."));
2150 return NULL;
2151 }
2152
2153 if (!dir->untracked->root) {
2154 const int len = sizeof(*dir->untracked->root);
2155 dir->untracked->root = xmalloc(len);
2156 memset(dir->untracked->root, 0, len);
2157 }
2158
2159 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2160 root = dir->untracked->root;
2161 if (hashcmp(dir->ss_info_exclude.sha1,
2162 dir->untracked->ss_info_exclude.sha1)) {
2163 invalidate_gitignore(dir->untracked, root);
2164 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2165 }
2166 if (hashcmp(dir->ss_excludes_file.sha1,
2167 dir->untracked->ss_excludes_file.sha1)) {
2168 invalidate_gitignore(dir->untracked, root);
2169 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2170 }
2171
2172 /* Make sure this directory is not dropped out at saving phase */
2173 root->recurse = 1;
2174 return root;
2175 }
2176
2177 int read_directory(struct dir_struct *dir, struct index_state *istate,
2178 const char *path, int len, const struct pathspec *pathspec)
2179 {
2180 struct untracked_cache_dir *untracked;
2181
2182 if (has_symlink_leading_path(path, len))
2183 return dir->nr;
2184
2185 untracked = validate_untracked_cache(dir, len, pathspec);
2186 if (!untracked)
2187 /*
2188 * make sure untracked cache code path is disabled,
2189 * e.g. prep_exclude()
2190 */
2191 dir->untracked = NULL;
2192 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2193 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2194 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2195 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2196
2197 /*
2198 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2199 * also pick up untracked contents of untracked dirs; by default
2200 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2201 */
2202 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2203 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2204 int i, j;
2205
2206 /* remove from dir->entries untracked contents of untracked dirs */
2207 for (i = j = 0; j < dir->nr; j++) {
2208 if (i &&
2209 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2210 FREE_AND_NULL(dir->entries[j]);
2211 } else {
2212 dir->entries[i++] = dir->entries[j];
2213 }
2214 }
2215
2216 dir->nr = i;
2217 }
2218
2219 if (dir->untracked) {
2220 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2221 trace_printf_key(&trace_untracked_stats,
2222 "node creation: %u\n"
2223 "gitignore invalidation: %u\n"
2224 "directory invalidation: %u\n"
2225 "opendir: %u\n",
2226 dir->untracked->dir_created,
2227 dir->untracked->gitignore_invalidated,
2228 dir->untracked->dir_invalidated,
2229 dir->untracked->dir_opened);
2230 if (dir->untracked == istate->untracked &&
2231 (dir->untracked->dir_opened ||
2232 dir->untracked->gitignore_invalidated ||
2233 dir->untracked->dir_invalidated))
2234 istate->cache_changed |= UNTRACKED_CHANGED;
2235 if (dir->untracked != istate->untracked) {
2236 FREE_AND_NULL(dir->untracked);
2237 }
2238 }
2239 return dir->nr;
2240 }
2241
2242 int file_exists(const char *f)
2243 {
2244 struct stat sb;
2245 return lstat(f, &sb) == 0;
2246 }
2247
2248 static int cmp_icase(char a, char b)
2249 {
2250 if (a == b)
2251 return 0;
2252 if (ignore_case)
2253 return toupper(a) - toupper(b);
2254 return a - b;
2255 }
2256
2257 /*
2258 * Given two normalized paths (a trailing slash is ok), if subdir is
2259 * outside dir, return -1. Otherwise return the offset in subdir that
2260 * can be used as relative path to dir.
2261 */
2262 int dir_inside_of(const char *subdir, const char *dir)
2263 {
2264 int offset = 0;
2265
2266 assert(dir && subdir && *dir && *subdir);
2267
2268 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2269 dir++;
2270 subdir++;
2271 offset++;
2272 }
2273
2274 /* hel[p]/me vs hel[l]/yeah */
2275 if (*dir && *subdir)
2276 return -1;
2277
2278 if (!*subdir)
2279 return !*dir ? offset : -1; /* same dir */
2280
2281 /* foo/[b]ar vs foo/[] */
2282 if (is_dir_sep(dir[-1]))
2283 return is_dir_sep(subdir[-1]) ? offset : -1;
2284
2285 /* foo[/]bar vs foo[] */
2286 return is_dir_sep(*subdir) ? offset + 1 : -1;
2287 }
2288
2289 int is_inside_dir(const char *dir)
2290 {
2291 char *cwd;
2292 int rc;
2293
2294 if (!dir)
2295 return 0;
2296
2297 cwd = xgetcwd();
2298 rc = (dir_inside_of(cwd, dir) >= 0);
2299 free(cwd);
2300 return rc;
2301 }
2302
2303 int is_empty_dir(const char *path)
2304 {
2305 DIR *dir = opendir(path);
2306 struct dirent *e;
2307 int ret = 1;
2308
2309 if (!dir)
2310 return 0;
2311
2312 while ((e = readdir(dir)) != NULL)
2313 if (!is_dot_or_dotdot(e->d_name)) {
2314 ret = 0;
2315 break;
2316 }
2317
2318 closedir(dir);
2319 return ret;
2320 }
2321
2322 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2323 {
2324 DIR *dir;
2325 struct dirent *e;
2326 int ret = 0, original_len = path->len, len, kept_down = 0;
2327 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2328 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2329 struct object_id submodule_head;
2330
2331 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2332 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2333 /* Do not descend and nuke a nested git work tree. */
2334 if (kept_up)
2335 *kept_up = 1;
2336 return 0;
2337 }
2338
2339 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2340 dir = opendir(path->buf);
2341 if (!dir) {
2342 if (errno == ENOENT)
2343 return keep_toplevel ? -1 : 0;
2344 else if (errno == EACCES && !keep_toplevel)
2345 /*
2346 * An empty dir could be removable even if it
2347 * is unreadable:
2348 */
2349 return rmdir(path->buf);
2350 else
2351 return -1;
2352 }
2353 strbuf_complete(path, '/');
2354
2355 len = path->len;
2356 while ((e = readdir(dir)) != NULL) {
2357 struct stat st;
2358 if (is_dot_or_dotdot(e->d_name))
2359 continue;
2360
2361 strbuf_setlen(path, len);
2362 strbuf_addstr(path, e->d_name);
2363 if (lstat(path->buf, &st)) {
2364 if (errno == ENOENT)
2365 /*
2366 * file disappeared, which is what we
2367 * wanted anyway
2368 */
2369 continue;
2370 /* fall thru */
2371 } else if (S_ISDIR(st.st_mode)) {
2372 if (!remove_dir_recurse(path, flag, &kept_down))
2373 continue; /* happy */
2374 } else if (!only_empty &&
2375 (!unlink(path->buf) || errno == ENOENT)) {
2376 continue; /* happy, too */
2377 }
2378
2379 /* path too long, stat fails, or non-directory still exists */
2380 ret = -1;
2381 break;
2382 }
2383 closedir(dir);
2384
2385 strbuf_setlen(path, original_len);
2386 if (!ret && !keep_toplevel && !kept_down)
2387 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2388 else if (kept_up)
2389 /*
2390 * report the uplevel that it is not an error that we
2391 * did not rmdir() our directory.
2392 */
2393 *kept_up = !ret;
2394 return ret;
2395 }
2396
2397 int remove_dir_recursively(struct strbuf *path, int flag)
2398 {
2399 return remove_dir_recurse(path, flag, NULL);
2400 }
2401
2402 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2403
2404 void setup_standard_excludes(struct dir_struct *dir)
2405 {
2406 dir->exclude_per_dir = ".gitignore";
2407
2408 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2409 if (!excludes_file)
2410 excludes_file = xdg_config_home("ignore");
2411 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2412 add_excludes_from_file_1(dir, excludes_file,
2413 dir->untracked ? &dir->ss_excludes_file : NULL);
2414
2415 /* per repository user preference */
2416 if (startup_info->have_repository) {
2417 const char *path = git_path_info_exclude();
2418 if (!access_or_warn(path, R_OK, 0))
2419 add_excludes_from_file_1(dir, path,
2420 dir->untracked ? &dir->ss_info_exclude : NULL);
2421 }
2422 }
2423
2424 int remove_path(const char *name)
2425 {
2426 char *slash;
2427
2428 if (unlink(name) && !is_missing_file_error(errno))
2429 return -1;
2430
2431 slash = strrchr(name, '/');
2432 if (slash) {
2433 char *dirs = xstrdup(name);
2434 slash = dirs + (slash - name);
2435 do {
2436 *slash = '\0';
2437 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2438 free(dirs);
2439 }
2440 return 0;
2441 }
2442
2443 /*
2444 * Frees memory within dir which was allocated for exclude lists and
2445 * the exclude_stack. Does not free dir itself.
2446 */
2447 void clear_directory(struct dir_struct *dir)
2448 {
2449 int i, j;
2450 struct exclude_list_group *group;
2451 struct exclude_list *el;
2452 struct exclude_stack *stk;
2453
2454 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2455 group = &dir->exclude_list_group[i];
2456 for (j = 0; j < group->nr; j++) {
2457 el = &group->el[j];
2458 if (i == EXC_DIRS)
2459 free((char *)el->src);
2460 clear_exclude_list(el);
2461 }
2462 free(group->el);
2463 }
2464
2465 stk = dir->exclude_stack;
2466 while (stk) {
2467 struct exclude_stack *prev = stk->prev;
2468 free(stk);
2469 stk = prev;
2470 }
2471 strbuf_release(&dir->basebuf);
2472 }
2473
2474 struct ondisk_untracked_cache {
2475 struct stat_data info_exclude_stat;
2476 struct stat_data excludes_file_stat;
2477 uint32_t dir_flags;
2478 unsigned char info_exclude_sha1[20];
2479 unsigned char excludes_file_sha1[20];
2480 char exclude_per_dir[FLEX_ARRAY];
2481 };
2482
2483 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2484 #define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
2485
2486 struct write_data {
2487 int index; /* number of written untracked_cache_dir */
2488 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2489 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2490 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2491 struct strbuf out;
2492 struct strbuf sb_stat;
2493 struct strbuf sb_sha1;
2494 };
2495
2496 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2497 {
2498 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2499 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2500 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2501 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2502 to->sd_dev = htonl(from->sd_dev);
2503 to->sd_ino = htonl(from->sd_ino);
2504 to->sd_uid = htonl(from->sd_uid);
2505 to->sd_gid = htonl(from->sd_gid);
2506 to->sd_size = htonl(from->sd_size);
2507 }
2508
2509 static void write_one_dir(struct untracked_cache_dir *untracked,
2510 struct write_data *wd)
2511 {
2512 struct stat_data stat_data;
2513 struct strbuf *out = &wd->out;
2514 unsigned char intbuf[16];
2515 unsigned int intlen, value;
2516 int i = wd->index++;
2517
2518 /*
2519 * untracked_nr should be reset whenever valid is clear, but
2520 * for safety..
2521 */
2522 if (!untracked->valid) {
2523 untracked->untracked_nr = 0;
2524 untracked->check_only = 0;
2525 }
2526
2527 if (untracked->check_only)
2528 ewah_set(wd->check_only, i);
2529 if (untracked->valid) {
2530 ewah_set(wd->valid, i);
2531 stat_data_to_disk(&stat_data, &untracked->stat_data);
2532 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2533 }
2534 if (!is_null_sha1(untracked->exclude_sha1)) {
2535 ewah_set(wd->sha1_valid, i);
2536 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2537 }
2538
2539 intlen = encode_varint(untracked->untracked_nr, intbuf);
2540 strbuf_add(out, intbuf, intlen);
2541
2542 /* skip non-recurse directories */
2543 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2544 if (untracked->dirs[i]->recurse)
2545 value++;
2546 intlen = encode_varint(value, intbuf);
2547 strbuf_add(out, intbuf, intlen);
2548
2549 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2550
2551 for (i = 0; i < untracked->untracked_nr; i++)
2552 strbuf_add(out, untracked->untracked[i],
2553 strlen(untracked->untracked[i]) + 1);
2554
2555 for (i = 0; i < untracked->dirs_nr; i++)
2556 if (untracked->dirs[i]->recurse)
2557 write_one_dir(untracked->dirs[i], wd);
2558 }
2559
2560 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2561 {
2562 struct ondisk_untracked_cache *ouc;
2563 struct write_data wd;
2564 unsigned char varbuf[16];
2565 int varint_len;
2566 size_t len = strlen(untracked->exclude_per_dir);
2567
2568 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2569 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2570 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2571 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2572 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2573 ouc->dir_flags = htonl(untracked->dir_flags);
2574
2575 varint_len = encode_varint(untracked->ident.len, varbuf);
2576 strbuf_add(out, varbuf, varint_len);
2577 strbuf_addbuf(out, &untracked->ident);
2578
2579 strbuf_add(out, ouc, ouc_size(len));
2580 FREE_AND_NULL(ouc);
2581
2582 if (!untracked->root) {
2583 varint_len = encode_varint(0, varbuf);
2584 strbuf_add(out, varbuf, varint_len);
2585 return;
2586 }
2587
2588 wd.index = 0;
2589 wd.check_only = ewah_new();
2590 wd.valid = ewah_new();
2591 wd.sha1_valid = ewah_new();
2592 strbuf_init(&wd.out, 1024);
2593 strbuf_init(&wd.sb_stat, 1024);
2594 strbuf_init(&wd.sb_sha1, 1024);
2595 write_one_dir(untracked->root, &wd);
2596
2597 varint_len = encode_varint(wd.index, varbuf);
2598 strbuf_add(out, varbuf, varint_len);
2599 strbuf_addbuf(out, &wd.out);
2600 ewah_serialize_strbuf(wd.valid, out);
2601 ewah_serialize_strbuf(wd.check_only, out);
2602 ewah_serialize_strbuf(wd.sha1_valid, out);
2603 strbuf_addbuf(out, &wd.sb_stat);
2604 strbuf_addbuf(out, &wd.sb_sha1);
2605 strbuf_addch(out, '\0'); /* safe guard for string lists */
2606
2607 ewah_free(wd.valid);
2608 ewah_free(wd.check_only);
2609 ewah_free(wd.sha1_valid);
2610 strbuf_release(&wd.out);
2611 strbuf_release(&wd.sb_stat);
2612 strbuf_release(&wd.sb_sha1);
2613 }
2614
2615 static void free_untracked(struct untracked_cache_dir *ucd)
2616 {
2617 int i;
2618 if (!ucd)
2619 return;
2620 for (i = 0; i < ucd->dirs_nr; i++)
2621 free_untracked(ucd->dirs[i]);
2622 for (i = 0; i < ucd->untracked_nr; i++)
2623 free(ucd->untracked[i]);
2624 free(ucd->untracked);
2625 free(ucd->dirs);
2626 free(ucd);
2627 }
2628
2629 void free_untracked_cache(struct untracked_cache *uc)
2630 {
2631 if (uc)
2632 free_untracked(uc->root);
2633 free(uc);
2634 }
2635
2636 struct read_data {
2637 int index;
2638 struct untracked_cache_dir **ucd;
2639 struct ewah_bitmap *check_only;
2640 struct ewah_bitmap *valid;
2641 struct ewah_bitmap *sha1_valid;
2642 const unsigned char *data;
2643 const unsigned char *end;
2644 };
2645
2646 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2647 {
2648 memcpy(to, data, sizeof(*to));
2649 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2650 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2651 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2652 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2653 to->sd_dev = ntohl(to->sd_dev);
2654 to->sd_ino = ntohl(to->sd_ino);
2655 to->sd_uid = ntohl(to->sd_uid);
2656 to->sd_gid = ntohl(to->sd_gid);
2657 to->sd_size = ntohl(to->sd_size);
2658 }
2659
2660 static int read_one_dir(struct untracked_cache_dir **untracked_,
2661 struct read_data *rd)
2662 {
2663 struct untracked_cache_dir ud, *untracked;
2664 const unsigned char *next, *data = rd->data, *end = rd->end;
2665 unsigned int value;
2666 int i, len;
2667
2668 memset(&ud, 0, sizeof(ud));
2669
2670 next = data;
2671 value = decode_varint(&next);
2672 if (next > end)
2673 return -1;
2674 ud.recurse = 1;
2675 ud.untracked_alloc = value;
2676 ud.untracked_nr = value;
2677 if (ud.untracked_nr)
2678 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2679 data = next;
2680
2681 next = data;
2682 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2683 if (next > end)
2684 return -1;
2685 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2686 data = next;
2687
2688 len = strlen((const char *)data);
2689 next = data + len + 1;
2690 if (next > rd->end)
2691 return -1;
2692 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2693 memcpy(untracked, &ud, sizeof(ud));
2694 memcpy(untracked->name, data, len + 1);
2695 data = next;
2696
2697 for (i = 0; i < untracked->untracked_nr; i++) {
2698 len = strlen((const char *)data);
2699 next = data + len + 1;
2700 if (next > rd->end)
2701 return -1;
2702 untracked->untracked[i] = xstrdup((const char*)data);
2703 data = next;
2704 }
2705
2706 rd->ucd[rd->index++] = untracked;
2707 rd->data = data;
2708
2709 for (i = 0; i < untracked->dirs_nr; i++) {
2710 len = read_one_dir(untracked->dirs + i, rd);
2711 if (len < 0)
2712 return -1;
2713 }
2714 return 0;
2715 }
2716
2717 static void set_check_only(size_t pos, void *cb)
2718 {
2719 struct read_data *rd = cb;
2720 struct untracked_cache_dir *ud = rd->ucd[pos];
2721 ud->check_only = 1;
2722 }
2723
2724 static void read_stat(size_t pos, void *cb)
2725 {
2726 struct read_data *rd = cb;
2727 struct untracked_cache_dir *ud = rd->ucd[pos];
2728 if (rd->data + sizeof(struct stat_data) > rd->end) {
2729 rd->data = rd->end + 1;
2730 return;
2731 }
2732 stat_data_from_disk(&ud->stat_data, rd->data);
2733 rd->data += sizeof(struct stat_data);
2734 ud->valid = 1;
2735 }
2736
2737 static void read_sha1(size_t pos, void *cb)
2738 {
2739 struct read_data *rd = cb;
2740 struct untracked_cache_dir *ud = rd->ucd[pos];
2741 if (rd->data + 20 > rd->end) {
2742 rd->data = rd->end + 1;
2743 return;
2744 }
2745 hashcpy(ud->exclude_sha1, rd->data);
2746 rd->data += 20;
2747 }
2748
2749 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2750 const unsigned char *data,
2751 const unsigned char *sha1)
2752 {
2753 stat_data_from_disk(&sha1_stat->stat, data);
2754 hashcpy(sha1_stat->sha1, sha1);
2755 sha1_stat->valid = 1;
2756 }
2757
2758 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2759 {
2760 struct untracked_cache *uc;
2761 struct read_data rd;
2762 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2763 const char *ident;
2764 int ident_len, len;
2765 const char *exclude_per_dir;
2766
2767 if (sz <= 1 || end[-1] != '\0')
2768 return NULL;
2769 end--;
2770
2771 ident_len = decode_varint(&next);
2772 if (next + ident_len > end)
2773 return NULL;
2774 ident = (const char *)next;
2775 next += ident_len;
2776
2777 if (next + ouc_size(0) > end)
2778 return NULL;
2779
2780 uc = xcalloc(1, sizeof(*uc));
2781 strbuf_init(&uc->ident, ident_len);
2782 strbuf_add(&uc->ident, ident, ident_len);
2783 load_sha1_stat(&uc->ss_info_exclude,
2784 next + ouc_offset(info_exclude_stat),
2785 next + ouc_offset(info_exclude_sha1));
2786 load_sha1_stat(&uc->ss_excludes_file,
2787 next + ouc_offset(excludes_file_stat),
2788 next + ouc_offset(excludes_file_sha1));
2789 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2790 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2791 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2792 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2793 next += ouc_size(strlen(exclude_per_dir));
2794 if (next >= end)
2795 goto done2;
2796
2797 len = decode_varint(&next);
2798 if (next > end || len == 0)
2799 goto done2;
2800
2801 rd.valid = ewah_new();
2802 rd.check_only = ewah_new();
2803 rd.sha1_valid = ewah_new();
2804 rd.data = next;
2805 rd.end = end;
2806 rd.index = 0;
2807 ALLOC_ARRAY(rd.ucd, len);
2808
2809 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2810 goto done;
2811
2812 next = rd.data;
2813 len = ewah_read_mmap(rd.valid, next, end - next);
2814 if (len < 0)
2815 goto done;
2816
2817 next += len;
2818 len = ewah_read_mmap(rd.check_only, next, end - next);
2819 if (len < 0)
2820 goto done;
2821
2822 next += len;
2823 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2824 if (len < 0)
2825 goto done;
2826
2827 ewah_each_bit(rd.check_only, set_check_only, &rd);
2828 rd.data = next + len;
2829 ewah_each_bit(rd.valid, read_stat, &rd);
2830 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2831 next = rd.data;
2832
2833 done:
2834 free(rd.ucd);
2835 ewah_free(rd.valid);
2836 ewah_free(rd.check_only);
2837 ewah_free(rd.sha1_valid);
2838 done2:
2839 if (next != end) {
2840 free_untracked_cache(uc);
2841 uc = NULL;
2842 }
2843 return uc;
2844 }
2845
2846 static void invalidate_one_directory(struct untracked_cache *uc,
2847 struct untracked_cache_dir *ucd)
2848 {
2849 uc->dir_invalidated++;
2850 ucd->valid = 0;
2851 ucd->untracked_nr = 0;
2852 }
2853
2854 /*
2855 * Normally when an entry is added or removed from a directory,
2856 * invalidating that directory is enough. No need to touch its
2857 * ancestors. When a directory is shown as "foo/bar/" in git-status
2858 * however, deleting or adding an entry may have cascading effect.
2859 *
2860 * Say the "foo/bar/file" has become untracked, we need to tell the
2861 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2862 * directory any more (because "bar" is managed by foo as an untracked
2863 * "file").
2864 *
2865 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2866 * was the last untracked entry in the entire "foo", we should show
2867 * "foo/" instead. Which means we have to invalidate past "bar" up to
2868 * "foo".
2869 *
2870 * This function traverses all directories from root to leaf. If there
2871 * is a chance of one of the above cases happening, we invalidate back
2872 * to root. Otherwise we just invalidate the leaf. There may be a more
2873 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2874 * detect these cases and avoid unnecessary invalidation, for example,
2875 * checking for the untracked entry named "bar/" in "foo", but for now
2876 * stick to something safe and simple.
2877 */
2878 static int invalidate_one_component(struct untracked_cache *uc,
2879 struct untracked_cache_dir *dir,
2880 const char *path, int len)
2881 {
2882 const char *rest = strchr(path, '/');
2883
2884 if (rest) {
2885 int component_len = rest - path;
2886 struct untracked_cache_dir *d =
2887 lookup_untracked(uc, dir, path, component_len);
2888 int ret =
2889 invalidate_one_component(uc, d, rest + 1,
2890 len - (component_len + 1));
2891 if (ret)
2892 invalidate_one_directory(uc, dir);
2893 return ret;
2894 }
2895
2896 invalidate_one_directory(uc, dir);
2897 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2898 }
2899
2900 void untracked_cache_invalidate_path(struct index_state *istate,
2901 const char *path)
2902 {
2903 if (!istate->untracked || !istate->untracked->root)
2904 return;
2905 invalidate_one_component(istate->untracked, istate->untracked->root,
2906 path, strlen(path));
2907 }
2908
2909 void untracked_cache_remove_from_index(struct index_state *istate,
2910 const char *path)
2911 {
2912 untracked_cache_invalidate_path(istate, path);
2913 }
2914
2915 void untracked_cache_add_to_index(struct index_state *istate,
2916 const char *path)
2917 {
2918 untracked_cache_invalidate_path(istate, path);
2919 }
2920
2921 /* Update gitfile and core.worktree setting to connect work tree and git dir */
2922 void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2923 {
2924 struct strbuf gitfile_sb = STRBUF_INIT;
2925 struct strbuf cfg_sb = STRBUF_INIT;
2926 struct strbuf rel_path = STRBUF_INIT;
2927 char *git_dir, *work_tree;
2928
2929 /* Prepare .git file */
2930 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
2931 if (safe_create_leading_directories_const(gitfile_sb.buf))
2932 die(_("could not create directories for %s"), gitfile_sb.buf);
2933
2934 /* Prepare config file */
2935 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
2936 if (safe_create_leading_directories_const(cfg_sb.buf))
2937 die(_("could not create directories for %s"), cfg_sb.buf);
2938
2939 git_dir = real_pathdup(git_dir_, 1);
2940 work_tree = real_pathdup(work_tree_, 1);
2941
2942 /* Write .git file */
2943 write_file(gitfile_sb.buf, "gitdir: %s",
2944 relative_path(git_dir, work_tree, &rel_path));
2945 /* Update core.worktree setting */
2946 git_config_set_in_file(cfg_sb.buf, "core.worktree",
2947 relative_path(work_tree, git_dir, &rel_path));
2948
2949 strbuf_release(&gitfile_sb);
2950 strbuf_release(&cfg_sb);
2951 strbuf_release(&rel_path);
2952 free(work_tree);
2953 free(git_dir);
2954 }
2955
2956 /*
2957 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
2958 */
2959 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
2960 {
2961 if (rename(old_git_dir, new_git_dir) < 0)
2962 die_errno(_("could not migrate git directory from '%s' to '%s'"),
2963 old_git_dir, new_git_dir);
2964
2965 connect_work_tree_and_git_dir(path, new_git_dir);
2966 }